Projecting future changes in potato yield using machine learning techniques: A case study for Prince Edward Island, Canada

Accurate prediction of crop yields is crucial for informed agricultural decision-making, ensuring food supply, and supporting farmers' livelihoods. While machine learning algorithms have been widely utilized for yield prediction in agricultural research, the full potential of capturing short and long-term dependencies has not been sufficiently explored in previous studies. This research demonstrates the efficacy of exploiting the nature of long and short-term dependency in modeling potato yield using machine learning techniques to facilitate agricultural planning and future projections. Precise models enable projecting crop yield under diverse climate scenarios, informing farmers about potential production changes, and capturing the interplay between climate factors and projected production at a regional level. The methodology employed in this study, encompassing data gathering, model selection, and projection creation, can be extended to other regions and crops. The projections for the end of the century in Prince Edward Island (PEI) reveal a substantial decline in potato yield under various scenarios. Specifically, under the SSP5-8.5 scenario, the projections suggest a potential decline of up to 70% in crop yield. In comparison, the SSP1 and SSP2 scenarios exhibit a relatively lower percentage decrease in crop yield, ranging from 4% to 15%. This highlights the importance of reducing greenhouse gas emissions to minimize the potential decline in potato yield. It also implies the need for introducing adaptive farming practices to ensure sustainable potato production in the context of climate change.